Today’s brief

October 8: the hybrid stack and the agent surge

Thursday, October 8, 20266 min read
Tools & tacticsThe one thing

Microsoft rebuilds Windows to run AI agents locally, not just in the cloud

At a San Francisco event, Microsoft unveiled a "hybrid intelligence" overhaul of Windows that embeds Copilot and autonomous AI agents directly into the OS, letting many tasks run on-device instead of in the cloud.

Why it matters

While Microsoft operates one of the industry's largest cloud computing services, the company is betting that having a strong presence throughout the "tech stack" could give it an edge as AI transforms the way businesses operate. Microsoft was one of the first Big Tech companies to recognize the importance of generative AI, investing $13 billion in OpenAI and partnering with the company years ago, but the alliance between the two companies weakened, and Microsoft is now trying to find its lane in a crowded AI landscape as competitors like Google, OpenAI, and Anthropic take the lead. This is Microsoft trying to win on the device itself, where it still has a near-monopoly, rather than on model quality, where it's fallen behind.

What this means for you

The redesigned Copilot app will now be a central interface through which customers can chat with AI bots, tap AI tools to help accomplish tasks, write code, and access a new Autopilot feature that brings agentic capabilities. That Autopilot feature will let users name their agent, assign it a role and a goal, and then set it loose on tasks such as project updates, tracking work decisions and deadlines, reminding colleagues when commitments are due, gathering prep materials, and reaching out to partners. Expect a standing, named agent quietly working through your inbox and files by default on the next Windows PC you're issued.

Engineers: A user will be able to have Copilot autonomously pull all of the files they need to file taxes from various parts of their laptop without accessing the cloud, making it cheaper and faster — "Copilot now truly understands your PC," according to Copilot head Jacob Andreou. Local inference means lower latency and no per-call cloud cost for routine tasks — useful to know when scoping what you build versus what Windows now does for free.

Managers: Microsoft is betting that this new "super app" will boost Copilot by connecting AI directly to its already lucrative business tools, racing to protect its turf as OpenAI and Anthropic push into enterprise offerings. In July, Microsoft 365 Copilot had surpassed 30 million paid seats, with net seats added doubling quarter over quarter — this is likely already on your procurement roadmap whether IT has flagged it or not.

Do this: If your org runs Windows and M365, ask IT/security now how agent permissions and data access will be scoped before Autopilot rolls out — don't wait for the default settings to land on your machine.

Signal 4/5· ImportantSource: Microsoft shows off new Windows software, revamped for agentic AI — Fortune

Mortgage startup Vesta raises $30M after letting AI agents make underwriting calls

Vesta, which automates mortgage loan origination with AI agents, raised a $30 million round led by Conversion Capital, with existing lender customers Pennymac and New American Funding joining as investors. The company says revenue is up 12x year over year as lenders move from supervised agents to ones that handle underwriting decisions on their own.

Why it matters & what to do
Why it matters

Mortgage underwriting is exactly the kind of workflow sceptics said agents couldn't touch — multi-stage, regulated, and expensive to get wrong at roughly $11,000 and 40 days per loan. CEO Mike Yu credits the jump in autonomy to a specific model upgrade, not just better prompting, which is a useful data point for anyone deciding whether to trust agents with real decisions.

What this means for you

If a compliance-heavy process like mortgage underwriting can be handed to agents — with humans approving first, then stepping back — your own "too risky for AI" workflow is probably next.

Finance: Lenders adopting this are reportedly cutting the labor cost that makes up most of that $11,000-per-loan origination bill, which is worth tracking if you cover or invest in mortgage or fintech names.

Managers: Vesta's rollout pattern — human approves every action, then the agent earns a growing share of cases on its own — is a practical template for introducing agents into any regulated or high-stakes team process.

Do this: If you manage a compliance-heavy workflow, ask whether a "human-approves-first, then expand autonomy" rollout could apply before assuming the process is off-limits to AI.

Source: Vesta raises $30M to bring swarms of agents to mortgage lenders — TechCrunch
Signal 3/5· Pay attention

Google's drug-discovery AI arm in talks to raise at $40 billion-plus valuation

Isomorphic Labs, the Google DeepMind spinout using AI for drug discovery, is in early talks to raise new funding at a valuation of at least $40 billion — and possibly as high as $50 billion.

Why it matters & what to do
Why it matters

Isomorphic raised a $2.1 billion round just five months ago; a jump to $40-50 billion would mark one of the fastest valuation climbs in the AI sector. It's also a signal that investors see more durable value in AI applied to a specific, high-stakes domain (biology) than in general-purpose chatbots.

What this means for you

Investors are increasingly betting that "AI plus a hard science" beats "AI alone" — the moat is in proprietary data and partnerships, not just model size.

Finance: A $40-50B valuation for a company with no drug yet in clinical trials is a bet on long-term licensing and milestone payments from partners like Eli Lilly and Novartis, not current revenue — classic frontier-biotech risk pricing.

Managers: If you work in a regulated, data-rich vertical (healthcare, law, engineering), this is a signal that building deep domain AI — not competing with OpenAI head-on — is where defensible value and funding are flowing.

Do this: Nothing to do yet — just note the shift: capital is rotating toward vertical AI applied to specific industries, not just general-purpose model makers.

Source: Alphabet's Isomorphic Labs in Funding Talks for at Least $40 Billion Value — Bloomberg
Signal 3/5· Pay attention

Nvidia-backed data center IPO stalls as investors balk at $30B price tag

Firmus Grid's planned $5.5 billion Australian IPO closed its order books without enough demand to clear the A$11 marketed price, putting the deal at risk of being pulled or repriced.

Why it matters & what to do
Why it matters

This is one of the first big AI infrastructure IPOs to hit real price resistance from outside investors, not just insiders rolling over stakes — a signal that public-market appetite for AI capex bets is thinner than the private-market hype suggested.

What this means for you

When a marquee, Nvidia-backed deal can't clear its price with real buyers, it's a early sign that capital allocators are getting pickier about AI infrastructure, not pulling back from AI altogether.

Finance: Watch where allocation ends up — if existing backers like Nvidia and Blackstone end up absorbing more of the deal to get it done, that's a weaker signal than genuine new-money demand, and a useful read on how much real conviction is behind AI data center valuations right now.

Do this: Nothing to do yet — just note this as an early read on AI infrastructure financing sentiment before the next big data center raise or IPO crosses your desk.

Sources: Nvidia-Backed IPO's Cratering Demand Sends Warning on AI Funding — Bloomberg, Nvidia-Backed Firmus Said to Close Books as IPO Struggles — Bloomberg
Signal 3/5· Pay attention

Sequoia leads $30M bet on AI agents that trade for you

Catalyst, a startup that turns plain-language investing intent into executed trades via AI agents, raised a $30 million seed round led by Sequoia, with Jump Trading, Coinbase, and Lux Capital also participating.

Why it matters & what to do
Why it matters

Catalyst's pilot already generated hundreds of millions in trading volume in a few weeks, and the product is now rolling out beyond power traders to everyday retail investors — a real-money test of whether people will hand trading decisions to an AI agent, not just drafting or research.

What this means for you

AI agents are moving from chat assistants into systems that hold your money and act on your behalf, with only a lightweight human "accept" step in between.

Finance: Expect more retail platforms to copy this "state your intent, agent executes" model, which will pressure traditional advisors and brokerages to automate or partner fast.

Do this: If you use any retail trading app, check whether it's adding agentic execution features and read the fine print on what "human in the loop" actually means before you opt in.

Source: Sequoia-backed Catalyst raises $30 million to build AI trading agents for everyday investors — Fortune
Signal 3/5· Pay attention
One line to sound smart

“Microsoft is building AI into Windows itself, Isomorphic Labs is raising at $40 billion on drug discovery, and AI agents are now handling mortgages and trades.”

Tool worth a look

Claude is the AI assistant this brief is built with — genuinely useful for drafting, summarizing dense material, and thinking through what a development actually means for you. An honest pick, not a paid link.

Try Claude →

Futureproof Daily is researched and written by AI against our editorial standards — see how we work. Sources are linked on each item. Nothing here is financial, investment, or legal advice.